Try proving a few more details. You can’t be vague with the new model, it’ll fallback to the basics it was trained in. And it’s obvious that the flex females predominantly had lip contouring.

You seem to have pretty good idea what you’re looking for. So, put it into words inside a prompt, and tune it phrase by phrase. The new model is quite good at catching the detailed description. The older model had its charm, but let’s be honest it was outdated.

check these out if they are any close to your idea, you could use my generator. Style -> Digital painting, filter: default (you can play with painting section filter, to an artist (scroll photo and anime artists to “Painting”) or a painting style (baroque, realism, etc) to try and tune it). Actor->Male, Person, 25, Masculine, Western European. Clothing -> Medieval armor. Camera -> artistic, facial expression: determined. Scene -> Forest, night, cloudy, dramatic. Once you find your thing, you can Tools -> Save this under a name. It’ll be available in Actor -> Main character -> Saved, so you don’t have to choose the settings every time. https://perchance.org/imageconstructor

Try describing it like this, “furry girl (boy/female/whatever), anthropomorphic animal with human female face, animal ears, animal tail, animal paws.” Animal can be any specific animal of your choice.

could you provide a link to your generator at least? for select elements id should always be different. if you want to use the same list twice you just change the id. but actually seeing what you are talking about would be a start.

I believe it’s kinda outdated. The new model prefers full sentences with details, rather than a word with brackets.

Okay, I might have missed something. I’ll have to investigate it further when it happens on animals and non-nudity scenes.

<input type="text" id="shownHere">
<select id="userChoosesHere" onchange="document.getElementById('shownHere').value=this.value;">

Something like this?

Btw, if you do Reset all, make sure to click “No”. It asks if it should also erase saved characters, that could also affect you loosing them.

It stores locally on your drive but I’ve been coding these days and changed a lot, that’s why it constantly resets for you. Sorry about that. Previously working version is not fully functional atm, because I added image dropdowns, and they are not saved atm. I’m gonna finalize the early beta soon. Until then, I’d recommend making a duplicate of the generator from the settings, so you’re not affected by me coding.

Just prohibit posting any nsfw stuff in the public library before logging in. That should fix most of the problems. The moderation system is not ideal and I assume it’s a lot of extra work.

I’m sure it’s just a part of a coding process. Give it some time. I also tried to make an icon of an armchair but it just wouldn’t show it without super detailed descriptions.

That could partially be mitigated in the original prompt. Try specifying each facial feature and how you wanna see it. I do something like “ultra detailed skin around eyes” etc when I’m focusing on it. Also, make sure to use 512x768 resolution, when you’re making a full body shot. This will prevent many distortion problems. Also, add “shallow depth of field lens” to focus more on the composition rather than background.

you can upscale images after generation, there are many AI plugins like that on the web. For example, https://www.aiease.ai/app/enhance-photo

flux.1-schnel

I’m open to any ideas, of course.

That’s my generator I was asking help for :)

Awesome! Thanks a lot!

That’s what CEO’s of major AI companies like OpenAI say on youtube and x. All but one of them are optimistic about continuous learning AI’s by mid 26, and the other guy says “within 5 years”.

The model is trained by the devs. Continuous learning from user experience is a thing of 2026 I think, maybe 2027.